I am happy to share another answer or two from our statistical consulting team to frequently asked questions (FAQs) about design of experiments (DOE), as well as timely alerts for events, publications, and software updates. Check it out! Feel free to get back to me via [email protected] with further questions or comments: I would really appreciate hearing from you!
Please do not send me requests to subscribe or unsubscribe, follow the instructions at the end of this message.
Sincerely, Mark J. Anderson, PE, CQE
Engineering Consultant, Stat-Ease, Inc.
PS Quote for the day: Words of wisdom from renowned statistician Sir David Cox, who passed away Jan 18 at the age of 97.
(Page down to the end of this e-zine to enjoy the actual quote.)
BLOGS StatsMadeEasy Blog
My wry look at all things statistical and/or scientific with an engineering perspective.
Also, see the Stat-Ease blogfor tips on making DOE easy. For example, a recent posting provides insights on “What’s Behind Aliasing in Fractional-Factorial Designs.” Take a look!
FAQ
How to Use Stat-Ease Software's Auto-Select tool Original question from a Lead Scientist: “I am using Design-Expert software’s Auto Select tool to fine-tune the predictive model for my most critical DOE response. Should I consider any other options to the default for AICc criterion with forward selection?”
Answer:
Good on you for not just accepting the results of this one approach to model reduction. Terms selected may vary somewhat based on the choice of criterion, the direction and threshold—especially for highly collinear inputs. After going forward with AICc (a measure of goodness of fit), reset your process order and try the backward selection. I suggest you then switch to the p-value criterion—my favorite for it being more familiar to non-statisticians—and again go backward, which I like for its inclusivity. Then pick the model that produces the best predicted R-squared and/or seems most sensible based on your subject-matter knowledge. As George Box said, all models are wrong, but some are useful: You must be the judge.
- Mark
PS: While preparing a case study for my February 2022 webinar on Evaluating and Exploiting Existing Data, I saw for myself how differing settings on Auto Select produce alternative models. Follow the link and watch the recording at the 36-minute mark to see the results from AICc forward versus p-value backward. One of the models proved most useful for our client.
PS Do you need a speaker on DOE for a learning session within your company or professional society at regional, national, or international levels? If so, please get back to me. – Mark
WORKSHOP ALERT Sharpen up on DOE—Enroll before classes fill
You can do no better for quickly advancing your DOE skills than attending a Stat-Ease workshop. Our expert instructors provide you with a lively and extremely informative series of lectures interspersed by valuable hands-on exercises. Enroll early to ensure your spot!